Remember that time you were over at your friend’s place? You saw a cool chair or table and asked them where they purchased it, only to find that they can’t remember? It’s a scenario that everyone has been in one time or another. What if you could finally get those answers by taking a simple photo of the furniture in question?
GrokStyle is a company that is working on an app that recognizes objects in a picture. Founded by Sean Bell and Kavita Bala, it uses a deep learning technology to learn visual similarities among products so it can find what you are looking for.
“GrokStyle basically gives you a buy button that answers questions like ‘I want that, what is it, where can I buy it?’ Then it also answers some follow-up questions like ‘what goes with that object and how do I use it in my home?’ We want to respond to all of these different questions using computer vision,” Bell told the Cornell Daily Sun. “You can take a picture on your phone or find it online and that will get into our system.”
The system works by using a deep neural network and algorithms to read an image. It processes it with many layers to predict a visual fingerprint. This fingerprint is nothing more than a list of numbers that represent vectors, but a large-scale search compares it with millions of other images that look similar. “Because we’re so accurate, the things that are most similar are mostly the same product,” Bell continued.
For now, GrokStyle will remain focused on identifying furniture due to its common structure and characteristics. In the future, Bell hopes to expand their outreach into other areas. Clothing is the logical next step, but that can be difficult since it changes shape when it is worn. This makes it difficult for computer vision. “We are planning on moving into fashion and starting with products where fit is less of an issue,” says Bell. “For example, handbags, shoes, and accessories.”
Currently, GrokStyle is still only available in demo form for select users, but people may contact the founders through their website.
